Original Research
The Law Firm AI Visibility Index
We scored 106 personal injury law firm websites across 12 major US metros on the 12 technical signals AI search engines use to find, trust, and cite a source. The verdict: most firms are findable, but not citable.
Last updated August 2, 2026. Methodology below.
Findable, and only sometimes citable
Nearly every firm we scored nails the fundamentals. 100% serve HTTPS, 99.1% have a title tag, and 96.2% ship at least some structured data. Traditional SEO has done its job: these sites are easy for a crawler to read.
The gap opens on the signals that decide whether an AI engine will actually cite a firm and attribute it correctly, and it is narrower than the first run suggested. 45.3% now expose attorney or author schema, the E-E-A-T signal that tells ChatGPT, Perplexity and Google AI Overviews who is behind the advice, so roughly half the market has that covered. Only 20.8% publish FAQ schema, the format AI assistants pull direct answers from, and that is the signal most firms are still missing. 64.2% have a social preview image.
Interestingly, 44.3% now serve an llms.txt file, mostly auto-generated by SEO plugins. That is the tell: firms are bolting on superficial AI signals without the underlying structure (answerable content, named experts, attribution) that earns citations. A plugin file is not a strategy.
How 106 firms scored, signal by signal
Amber bars mark signals present on fewer than 60% of firms. These are the AI-visibility gaps.
Score distribution
Methodology
We sampled 106 personal injury law firm websites across 12 major US metros (Atlanta, Dallas, Chicago, Phoenix, New York City, Los Angeles, Houston, Miami, Philadelphia, Denver, Seattle, Boston), sourced from public search results. Each homepage was fetched and scored out of 100 across 12 signals grouped into Technical Foundation, Content Signals, Structured Data, and AI Crawlability, weighted toward the structured-data and crawlability signals AI engines rely on most. Detection is purely technical (no AI-generated estimates): we parse the page HTML, JSON-LD, llms.txt, and robots.txt directly. llms.txt and robots.txt are only counted when served as plain text, to avoid false positives from sites that return their homepage for unknown paths. The scan runs on the same engine as our free checker, and the figures below are from the August 2, 2026 run under the current scoring. That scoring was tightened after the first published run: an unreadable robots.txt no longer counts as crawler access, llms.txt carries less weight, Organization and author credit require a named entity rather than a matching type string, and FAQ credit requires at least one question to be visible on the page. The parser was also rewritten to walk every node in a JSON-LD block instead of only the root and its @graph array. That last change matters for comparison: the first run reported 17.9% author schema and a median of 70, but it was missing entities nested inside other nodes. The difference between the two runs is mostly a corrected instrument rather than eight weeks of market movement.
Where does your firm score?
Run the same check on your own site in seconds. Free, no signup, instant breakdown of what to fix first.